Photo by LlamaIndex Blog, May 7

Latest News

  1. Google shows Med-Gemini with some great results
    This is the new family of multimodal models for medicine unlocking new possibilities for health. It offers more accurate multimodal conversations about medical images, surgical videos, genomics, health records, ECGs with SOTA benchmark results.

  2. Gradient releases 1M context window LLama-3 finetune
    The first LLama-3 8B with a context length of over 1M on Hugging Face- coming off the 160K context length model

  3. Stack Overflow and OpenAI Partner to Strengthen the World’s Most Popular Large Language Models
    OpenAI will surface validated technical knowledge from Stack Overflow directly into ChatGPT, giving users easy access to trusted, attributed, accurate, and highly technical knowledge and code backed by the millions of developers that have contributed to the Stack Overflow platform for 15 years

  4. Mysterious “gpt2-chatbot” AI model appears suddenly, confuses experts
    A newly surfaced AI model dubbed "gpt2-chatbot" has abruptly appeared, stirring confusion among experts. Its emergence on the LMSYS Chatbot Arena has sparked conversations within the AI community regarding its possible connection to upcoming OpenAI models. Despite showcasing commendable performance, evaluations suggest it falls short of surpassing GPT-4, leaving uncertainties lingering about its precise origins and specifications.

  5. Claude releases IOS app and Team accounts
    The power of frontier intelligence is now in your back pocket.

  6. GitHub Copilot Workspace
    GitHub introduces Copilot Workspace, an all-encompassing developer environment designed to streamline the coding journey from planning to deployment. Through intuitive natural language commands, this tool empowers AI industry professionals with a unified solution for enhancing development workflows.

  7. OpenAI CEO Sam Altman says GPT-4 is the dumbest AI model you’ll ever have to use again
    Sam Altman of OpenAI views GPT-4 as merely the tip of the iceberg in the company's journey towards more advanced models like GPT-5. Anticipated to possess upgraded capabilities such as video generation, GPT-5 heralds a future where AI evolves into exceptionally efficient aides, effortlessly executing tasks and offering solutions.

  8. A ChatGPT search engine is rumored to be coming next week
    There are rumors circulating about OpenAI's purported launch of a search engine based on ChatGPT, possibly under the domain "search.chatgpt.com." This move is speculated to be aimed at competing with Google by blending a chatbot feature with conventional search outcomes. Such a development mirrors the industry's inclination towards AI potentially transforming conventional web search practices.

  9. Pro music video powered by OpenAI’s Sora released in a world- first
    Under the direction of Paul Trillo, the official music video for Washed Out's "The Hardest Part" was crafted utilizing OpenAI's Sora, a text-to-video AI. This endeavor resulted in the creation of 700 clips, with only 55 making the final cut. The project has sparked ethical dialogues within the AI industry.

Articles

  1. Comparison of Llama-3 and Phi-3 using RAG
  2. Run Llama 3 with PandasAI and Ollama Locally
  3. Fine-tuning Llama 3: 70B for Code-Related Tasks
  4. OpenAI Is Doomed? - Et tu, Microsoft?
  5. How does temperature impact next token prediction in LLMs?
  6. Microsoft’s New AI Powerhouse: Unveiling MAI-1 and the Race for Language Model Supremacy

Papers and Repositories

  1. Vibes-Eval from Reka - a set of multimodal evals [Announcement, Paper, HF dataset
  2. WorldGPT: Empowering LLM as Multimodal World Model [Paper, Code]
  3. RAG and RAU: A Survey on Retrieval-Augmented Language Model in Natural Language Processing [Paper, Code]
  4. HuggingFace FINEWEB - a massive 45TB (the GPT4 of datasets) and 15T tokens high quality web data dataset [HF]